arXiv subjects

Bach Nguyen

Publications and source records attributed to Bach Nguyen.

15 recordsLinked to original sources

arXiv2026

VisAnalog: A Diagnostic Suite for Visual Concept Transfer on Natural Images

A useful test of visual concept learning is not just whether a model can recognize a concept in a single image, but whether it can preserve and manipulate concept-level properties under transformation and transfer them to new scenes. We introduce VisAnalog, a controlled suite for this setting on natural images. Each example instantiates $A\!:\!B::C\!:\,?$: images $B$ and a hidden target image $D$ are produced by applying the same deterministic transformation sequence to source images $A$ and $C$. Given $A$, $B$, and $C$, a model must answer a multiple-choice question about $D$. The benchmark contains 617 human-validated questions spanning one- to four-step transformations such as zoom, quadrant swap, rotation, flip, and hue rotation. Across strong proprietary and open-source VLMs, end-to-end accuracy is substantially lower than oracle accuracy when $D$ is directly shown, and degrades sharply as transformation depth increases, while human performance remains near the ceiling. A program-conditioned evaluation further separates failures of relation inference from failures of transformation application, showing that inferring the visual relation from $A \rightarrow B$ is the dominant bottleneck, with additional application errors emerging on harder multi-step cases. The dataset is publicly available at https://huggingface.co/datasets/zli99/VisAnalog.

cs.CV↗

arXiv2026

Seed Layer Engineering for Effective Charge Transfer Doping of MoS$_2$ Transistors

Integrating two-dimensional semiconductors such as MoS$_2$ with dielectric materials remains a central challenge for their use in future logic technologies. While seed layers are typically introduced to promote dielectric nucleation and adhesion, we show that they also critically govern charge-transfer doping and, in turn, transistor performance. Back-gated monolayer MoS$_2$ transistors passivated on their top-surface with a Ta-seed/HfO$_x$ dielectric stack were fabricated and characterized electrically and physically using Raman, photoluminescence, and X-ray photoelectron spectroscopies. Threshold voltage and on-current varied strongly with Ta-seed thickness and deposition conditions, and these changes correlated with signatures observed across all spectroscopic probes. The results reveal that the seed layer both introduces disorder into the MoS$_2$ channel and modifies the interfacial charge environment controlling charge transfer between HfO$_x$ and MoS$_2$. Optical spectroscopy shows that on-current tracks seed-induced disorder, whereas X-ray photoelectron spectroscopy indicates that threshold voltage correlates with shifts in the local electrostatic environment associated with interfacial charge transfer. Better performance was obtained with ultrathin 0.2 nm Ta seed layers deposited under oxygen-poor conditions, which limit deposition-induced damage while facilitating charge transfer. These findings identify seed-layer engineering as a key strategy for controlling disorder and interfacial doping in MoS$_2$ devices and establish multimodal spectroscopy as a practical during-fabrication approach for process development and monitoring.

cond-mat.mtrl-sci↗

arXiv2024

Accessing the topological properties of human brain functional sub-circuits in Echo State Networks

Recent years have witnessed an emerging trend in neuromorphic computing that centers around the use of brain connectomics as a blueprint for artificial neural networks. Connectomics-based neuromorphic computing has primarily focused on embedding human brain large-scale structural connectomes (SCs), as estimated from diffusion Magnetic Resonance Imaging (dMRI) modality, to echo-state networks (ESNs). A critical step in ESN embedding requires pre-determined read-in and read-out layers constructed by the induced subgraphs of the embedded reservoir. As \textit{a priori} set of functional sub-circuits are derived from functional MRI (fMRI) modality, it is unknown, till this point, whether the embedding of fMRI-induced sub-circuits/networks onto SCs is well justified from the neuro-physiological perspective and ESN performance across a variety of tasks. This paper proposes a pipeline to implement and evaluate ESNs with various embedded topologies and processing/memorization tasks. To this end, we showed that different performance optimums highly depend on the neuro-physiological characteristics of these pre-determined fMRI-induced sub-circuits. In general, fMRI-induced sub-circuit-embedded ESN outperforms simple bipartite and various null models with feed-forward properties commonly seen in MLP for different tasks and reservoir criticality conditions. We provided a thorough analysis of the topological properties of pre-determined fMRI-induced sub-circuits and highlighted their graph-theoretical properties that play significant roles in determining ESN performance.

q-bio.NC↗

arXiv2024

Investigating Context-Faithfulness in Large Language Models: The Roles of Memory Strength and Evidence Style

Retrieval-augmented generation (RAG) improves Large Language Models (LLMs) by incorporating external information into the response generation process. However, how context-faithful LLMs are and what factors influence LLMs' context faithfulness remain largely unexplored. In this study, we investigate the impact of memory strength and evidence presentation on LLMs' receptiveness to external evidence. We quantify the memory strength of LLMs by measuring the divergence in LLMs' responses to different paraphrases of the same question, which is not considered by previous works. We also generate evidence in various styles to examine LLMs' behavior. Our results show that for questions with high memory strength, LLMs are more likely to rely on internal memory. Furthermore, presenting paraphrased evidence significantly increases LLMs' receptiveness compared to simple repetition or adding details. These findings provide key insights for improving retrieval-augmented generation and context-aware LLMs. Our code is available at https://github.com/liyp0095/ContextFaithful.

cs.CL↗

arXiv2024

A Consistent Lebesgue Measure for Multi-label Learning

Multi-label loss functions are usually non-differentiable, requiring surrogate loss functions for gradient-based optimisation. The consistency of surrogate loss functions is not proven and is exacerbated by the conflicting nature of multi-label loss functions. To directly learn from multiple related, yet potentially conflicting multi-label loss functions, we propose a Consistent Lebesgue Measure-based Multi-label Learner (CLML) and prove that CLML can achieve theoretical consistency under a Bayes risk framework. Empirical evidence supports our theory by demonstrating that: (1) CLML can consistently achieve state-of-the-art results; (2) the primary performance factor is the Lebesgue measure design, as CLML optimises a simpler feedforward model without additional label graph, perturbation-based conditioning, or semantic embeddings; and (3) an analysis of the results not only distinguishes CLML's effectiveness but also highlights inconsistencies between the surrogate and the desired loss functions.

cs.LG↗

arXiv2023

Explicit constructions of connections on the projective line with a maximally ramified irregular singularity

The Deligne--Simpson problem is an existence problem for connections with specified local behavior. Almost all previous work on this problem has restricted attention to connections with regular or unramified singularities. Recently, the authors, together with Kulkarni and Matherne, formulated a version of the Deligne--Simpson problem where certain ramified singular points are allowed and solved it for the case of Coxeter connections, i.e., connections on the Riemann sphere with a maximally ramified singularity at zero and (possibly) an additional regular singular point at infinity. A certain matrix completion problem, which we call the Upper Nilpotent Completion Problem, plays a key role in our solution. This problem was solved by Krupnik and Leibman, but their work does not provide a practical way of constructing explicit matrix completions. Accordingly, our previous work does not give explicit Coxeter connections with specified singularities. In this paper, we provide a numerically stable and highly efficient algorithm for producing upper nilpotent completions of certain matrices that arise in the theory of Coxeter connections. Moreover, we show how the matrices generated by this algorithm can be used to provide explicit constructions of Coxeter connections with arbitrary unipotent monodromy in each case that such a connection exists.

math.AG↗

arXiv2022

The Software Stack That Won the Formula Student Driverless Competition

This report describes our approach to design and evaluate a software stack for a race car capable of achieving competitive driving performance in the different disciplines of the Formula Student Driverless. By using a 360{\deg} LiDAR and optionally three cameras, we reliably recognize the plastic cones that mark the track boundaries at distances of around 35 m, enabling us to drive at the physical limits of the car. Using a GraphSLAM algorithm, we are able to map these cones with a root-mean-square error of less than 15 cm while driving at speeds of over 70 kph on a narrow track. The high-precision map is used in the trajectory planning to detect the lane boundaries using Delaunay triangulation and a parametric cubic spline. We calculate an optimized trajectory using a minimum curvature approach together with a GGS-diagram that takes the aerodynamics at different velocities into account. To track the target path with accelerations of up to 1.6 g, the control system is split into a PI controller for longitudinal control and model predictive controller for lateral control. Additionally, a low-level optimal control allocation is used. The software is realized in ROS C++ and tested in a custom simulation, as well as on the actual race track.

cs.RO↗

arXiv2022

Poisson geometry and Azumaya loci of cluster algebras

There are two main types of objects in the theory of cluster algebras: the upper cluster algebras ${{\boldsymbol{\mathsf U}}}$ with their Gekhtman-Shapiro-Vainshtein Poisson brackets and their root of unity quantizations ${{\boldsymbol{\mathsf U}}}_\varepsilon$. On the Poisson side, we prove that (without any assumptions) the spectrum of every finitely generated upper cluster algebra ${{\boldsymbol{\mathsf U}}}$ with its GSV Poisson structure always has a Zariski open orbit of symplectic leaves and give an explicit description of it. On the quantum side, we describe the fully Azumaya loci of the quantizations ${{\boldsymbol{\mathsf U}}}_\varepsilon$ under the assumption that ${{\boldsymbol{\mathsf A}}}_\varepsilon = {{\boldsymbol{\mathsf U}}}_\varepsilon$ and ${{\boldsymbol{\mathsf U}}}_\varepsilon$ is a finitely generated algebra. All results allow frozen variables to be either inverted or not.

math.RT↗

arXiv2021

The Deligne-Simpson problem for connections on $\mathbb{G}_m$ with a maximally ramified singularity

The classical additive Deligne-Simpson problem is the existence problem for Fuchsian connections with residues at the singular points in specified adjoint orbits. Crawley-Boevey found the solution in 2003 by reinterpreting the problem in terms of quiver varieties. A more general version of this problem, solved by Hiroe, allows additional unramified irregular singularities. We apply the theory of fundamental and regular strata due to Bremer and Sage to formulate a version of the Deligne-Simpson problem in which certain ramified singularities are allowed. These allowed singular points are called toral singularities; they are singularities whose leading term with respect to a lattice chain filtration is regular semisimple. We solve this problem in the important special case of connections on $\mathbb{G}_m$ with a maximally ramified singularity at $0$ and possibly an additional regular singular point at infinity. We also give a complete characterization of all such connections which are rigid, under the additional hypothesis of unipotent monodromy at infinity.

math.AG↗

arXiv2020

Root of unity quantum cluster algebras and discriminants

We describe a connection between the subjects of cluster algebras, polynomial identity algebras and discriminants. For this, we define the notion of root of unity quantum cluster algebras and prove that they are polynomial identity algebras. Inside each such algebra we construct a (large) canonical central subalgebra, which can be viewed as a far reaching generalization of the central subalgebras of big quantum groups constructed by De Concini, Kac and Procesi and used in representation theory. Each such central subalgebra is proved to be isomorphic to the underlying classical cluster algebra of geometric type. When the root of unity quantum cluster algebra is free over its central subalgebra, we prove that the discriminant of the pair is a product of powers of the frozen variables times an integer. An extension of this result is also proved for the discriminants of all subalgebras generated by the cluster variables of nerves in the exchange graph. These results can be used for the effective computation of discriminants. As an application we prove an explicit formula for the discriminant of the integral form over ${\mathbb{Z}}[\varepsilon]$ of each quantum unipotent cells of De Concini, Kac and Procesi for arbitrary symmetrizable Kac-Moody algebras, where $\varepsilon$ is a root of unity.

math.QA↗

arXiv2019

Actions of cocommutative Hopf algebras

Let $H$ be a cocommutative Hopf algebra acting on an algebra $A$. Assuming the base field to be algebraically closed and the $H$-action on $A$ to be integral, that is, it is given by a coaction of some Hopf subalgebra of the finite dual $H^\circ$ that is an integral domain, we stratify the prime spectrum $\mbox{Spec}\, A$ in terms of the prime spectra of certain commutative algebras. For arbitrary $H$-actions in characteristic $0$, we show that the largest $H$-stable ideal of $A$ that is contained in a given semiprime ideal of $A$ is semiprime as well.

math.RA↗

arXiv2019

On the Adjoint Representation of a Hopf Algebra

We consider the adjoint representation of a Hopf algebra $H$ focusing on the locally finite part, $H_{\text{adfin}}$, defined as the sum of all finite-dimensional subrepresentations. For virtually cocommutative $H$ (i.e., $H$ is finitely generated as module over a cocommutative Hopf subalgebra), we show that $H_{\text{adfin}}$ is a Hopf subalgebra of $H$. This is a consequence of the fact, proved here, that locally finite parts yield a tensor functor on the module category of any virtually pointed Hopf algebra. For general Hopf algebras, $H_{\text{adfin}}$ is shown to be a left coideal subalgebra. We also prove a version of Dietzmann's Lemma from group theory for Hopf algebras.

math.RT↗

arXiv2018

Noncommutative discriminants via Poisson primes

We present a general method for computing discriminants of noncommutative algebras. It builds a connection with Poisson geometry and expresses the discriminants as products of Poisson primes. The method is applicable to algebras obtained by specialization from families, such as quantum algebras at roots of unity. It is illustrated with the specializations of the algebras of quantum matrices at roots of unity and more generally all quantum Schubert cell algebras.

math.RA↗

arXiv2017

Irreducibility of the Cayley-Menger determinant, and of a class of related polynomials

If $S$ is a given regular $n$-simplex, $n \ge 2$, of edge length $a$, then the distances $a_1$, $\cdots$, $a_{n+1}$ of an arbitrary point in its affine hull to its vertices are related by the fairly known elegant relation $ϕ_{n+1} (a,a_1,\cdots,a_{n+1})=0$, where $$ϕ= ϕ_t (x, x_1,\cdots,x_{n+1}) = \left( x^2+x_1^2+\cdots+x_{n+1}^2\right)^2 - t\left( x^4+x_1^4+\cdots+x_{n+1}^4\right).$$ The natural question whether this is essentially the only relation is answered positively by M. Hajja, M. Hayajneh, B. Nguyen, and Sh. Shaqaqha in a recently submitted paper entitled "Distances from the vertices of a regular simplex." In that paper, the authors made use of the irreducibility of the polynomial $ϕ$ in the case when $n \ge 2$, $t=n+1$, $x= a \ne 0$, and $k = \mathbb{R}$, but supplied no proof, promising to do so in another paper that is turning out to be this one. It is thus the main aim of this paper to establish that irreducibility. In fact, we treat the irreducibility of $ϕ$ without restrictions on $t$, $x$, $a$, and $k$. As a by-product, we obtain new proofs of results pertaining to the irreducibility of the general Cayley-Menger determinant that are more general than those established by C. D'Andrea and M. Sombra in Sib. J. Math. 46, 71--76.

math.AC↗

arXiv2016

Distances from the vertices of a regular simplex

If $S$ is a given regular $d$-simplex of edge length $a$ in the $d$-dimensional Euclidean space $\mathcal{E}$, then the distances $t_1$, $\cdots$, $t_{d+1}$ of an arbitrary point in $\mathcal{E}$ to the vertices of $S$ are related by the elegant relation $$(d+1)\left( a^4+t_1^4+\cdots+t_{d+1}^4\right)=\left( a^2+t_1^2+\cdots+t_{d+1}^2\right)^2.$$ The purpose of this paper is to prove that this is essentially the only relation that exists among $t_1,\cdots,t_{d+1}.$ The proof uses tools from analysis, algebra, and geometry.

math.MG↗